A conditional mutual information based selectional association and word sense disambiguation

This paper presents a conditional mutual information based selectional association to measure the selectional preference between two words in the same sentence. This selectional association is integrated conditional mutual information and a syntactic knowledge called link grammar. The selectional association is applied to indicative words selection for target words disambiguation. The experimental results show that this conditional mutual information based selectional association is able to select the appropriate word to indicate the appropriate meaning to the target word in different context.